QoS Routing Using Traffic Forecast - A Case Study of Time-Dependent Routing Yuekang Yang Chung-Horng Lung Dept. of Systems and Computer Engineering, Carleton.

Slides:



Advertisements
Similar presentations
Ch. 12 Routing in Switched Networks
Advertisements

Stable Load Control with Load Prediction in Multipath Packet Forwarding IlKyu Park, Youngseok Lee, and Yanghee Choi Proc. 15 th IEEE Int l conf. on Information.
Quality-of-Service Routing in IP Networks Donna Ghosh, Venkatesh Sarangan, and Raj Acharya IEEE TRANSACTIONS ON MULTIMEDIA JUNE 2001.
Ch. 12 Routing in Switched Networks Routing in Packet Switched Networks Routing Algorithm Requirements –Correctness –Simplicity –Robustness--the.
Data and Computer Communications
Hadi Goudarzi and Massoud Pedram
Shi Bai, Weiyi Zhang, Guoliang Xue, Jian Tang, and Chonggang Wang University of Minnesota, AT&T Lab, Arizona State University, Syracuse University, NEC.
© 2006 Cisco Systems, Inc. All rights reserved. MPLS v2.2—8-1 MPLS TE Overview Understanding MPLS TE Components.
© 2006 Cisco Systems, Inc. All rights reserved. MPLS v2.2—8-1 MPLS TE Overview Introducing the TE Concept.
1 Traffic Engineering (TE). 2 Network Congestion Causes of congestion –Lack of network resources –Uneven distribution of traffic caused by current dynamic.
Designing a New Routing Simulator for DiffServ MPLS Networks Peng Zhang Zhansong Ma Raimo Kantola {pgzhang, zhansong,
1 EL736 Communications Networks II: Design and Algorithms Class3: Network Design Modeling Yong Liu 09/19/2007.
A Flexible Model for Resource Management in Virtual Private Networks Presenter: Huang, Rigao Kang, Yuefang.
1 EL736 Communications Networks II: Design and Algorithms Class8: Networks with Shortest-Path Routing Yong Liu 10/31/2007.
A Case for Relative Differentiated Services and the Proportional Differentiation Model Constantinos Dovrolis Parameswaran Ramanathan University of Wisconsin-Madison.
AHOP Problem and QoS Route Pre-computation Adam Sachitano IAL.
Ashish Gupta Under Guidance of Prof. B.N. Jain Department of Computer Science and Engineering Advanced Networking Laboratory.
Volcano Routing Scheme Routing in a Highly Dynamic Environment Yashar Ganjali Stanford University Joint work with: Nick McKeown SECON 2005, Santa Clara,
Traffic Engineering Jennifer Rexford Advanced Computer Networks Tuesdays/Thursdays 1:30pm-2:50pm.
Multiple constraints QoS Routing Given: - a (real time) connection request with specified QoS requirements (e.g., Bdw, Delay, Jitter, packet loss, path.
Detecting Network Intrusions via Sampling : A Game Theoretic Approach Presented By: Matt Vidal Murali Kodialam T.V. Lakshman July 22, 2003 Bell Labs, Lucent.
Path Protection in MPLS Networks Ashish Gupta Design and Evaluation of Fault Tolerance Algorithms with Performance Constraints.
December 20, 2004MPLS: TE and Restoration1 MPLS: Traffic Engineering and Restoration Routing Zartash Afzal Uzmi Computer Science and Engineering Lahore.
1IMIC, 8/30/99 Constraint-Based Unicast and Multicast: Practical Issues Bala Rajagopalan NEC C&C Research Labs Princeton, NJ
Spring Routing & Switching Umar Kalim Dept. of Communication Systems Engineering 17/04/2007.
Path Protection in MPLS Networks Using Segment Based Approach.
Dynamic routing – QoS routing Load sensitive routing QoS routing.
A General approach to MPLS Path Protection using Segments Ashish Gupta Ashish Gupta.
S. Suri, M, Waldvogel, P. Warkhede CS University of Washington Profile-Based Routing: A New Framework for MPLS Traffic Engineering.
On Self Adaptive Routing in Dynamic Environments -- A probabilistic routing scheme Haiyong Xie, Lili Qiu, Yang Richard Yang and Yin Yale, MR and.
Jennifer Rexford Princeton University MW 11:00am-12:20pm Wide-Area Traffic Management COS 597E: Software Defined Networking.
1 Algorithms for Bandwidth Efficient Multicast Routing in Multi-channel Multi-radio Wireless Mesh Networks Hoang Lan Nguyen and Uyen Trang Nguyen Presenter:
SMUCSE 8344 Constraint-Based Routing in MPLS. SMUCSE 8344 Constraint Based Routing (CBR) What is CBR –Each link a collection of attributes (performance,
MATE: MPLS Adaptive Traffic Engineering Anwar Elwalid, et. al. IEEE INFOCOM 2001.
Distributed Quality-of-Service Routing of Best Constrained Shortest Paths. Abdelhamid MELLOUK, Said HOCEINI, Farid BAGUENINE, Mustapha CHEURFA Computers.
1 Meeyoung Cha (KAIST) Sue Moon (KAIST) Chong-Dae Park (KAIST) Aman Shaikh (AT&T Labs – Research) IEEE INFOCOM 2005 Poster Session Positioning Relay Nodes.
Cost-Performance Tradeoffs in MPLS and IP Routing Selma Yilmaz Ibrahim Matta Boston University.
1 Multi Protocol Label Switching Presented by: Petros Ioannou Dept. of Electrical and Computer Engineering, UCY.
1 Meeyoung Cha, Sue Moon, Chong-Dae Park Aman Shaikh Placing Relay Nodes for Intra-Domain Path Diversity To appear in IEEE INFOCOM 2006.
Network Aware Resource Allocation in Distributed Clouds.
“Intra-Network Routing Scheme using Mobile Agents” by Ajay L. Thakur.
An Integration Framework for Sensor Networks and Data Stream Management Systems.
MPLS and Traffic Engineering Ji-Hoon Yun Computer Communications and Switching Systems Lab.
Improving QoS Support in Mobile Ad Hoc Networks Agenda Motivations Proposed Framework Packet-level FEC Multipath Routing Simulation Results Conclusions.
2007/03/26OPLAB, NTUIM1 A Proactive Tree Recovery Mechanism for Resilient Overlay Network Networking, IEEE/ACM Transactions on Volume 15, Issue 1, Feb.
1 A Framework for Measuring and Predicting the Impact of Routing Changes Ying Zhang Z. Morley Mao Jia Wang.
Intradomain Traffic Engineering By Behzad Akbari These slides are based in part upon slides of J. Rexford (Princeton university)
1 - CS7701 – Fall 2004 Review of: Detecting Network Intrusions via Sampling: A Game Theoretic Approach Paper by: – Murali Kodialam (Bell Labs) – T.V. Lakshman.
6 December On Selfish Routing in Internet-like Environments paper by Lili Qiu, Yang Richard Yang, Yin Zhang, Scott Shenker presentation by Ed Spitznagel.
High-Speed Policy-Based Packet Forwarding Using Efficient Multi-dimensional Range Matching Lakshman and Stiliadis ACM SIGCOMM 98.
On Reducing Mesh Delay for Peer- to-Peer Live Streaming Dongni Ren, Y.-T. Hillman Li, S.-H. Gary Chan Department of Computer Science and Engineering The.
Supporting DiffServ with Per-Class Traffic Engineering in MPLS.
1 An Arc-Path Model for OSPF Weight Setting Problem Dr.Jeffery Kennington Anusha Madhavan.
Introducing a New Concept in Networking Fluid Networking S. Wood Nov Copyright 2006 Modern Systems Research.
1 Slides by Yong Liu 1, Deep Medhi 2, and Michał Pióro 3 1 Polytechnic University, New York, USA 2 University of Missouri-Kansas City, USA 3 Warsaw University.
Efficient Resource Allocation for Wireless Multicast De-Nian Yang, Member, IEEE Ming-Syan Chen, Fellow, IEEE IEEE Transactions on Mobile Computing, April.
Routing in Delay Tolerant Network Qing Ye EDIFY Group of Lehigh University.
1 Traffic Engineering By Kavitha Ganapa. 2 Introduction Traffic engineering is concerned with the issue of performance evaluation and optimization of.
Internet Traffic Engineering Motivation: –The Fish problem, congested links. –Two properties of IP routing Destination based Local optimization TE: optimizing.
Placing Relay Nodes for Intra-Domain Path Diversity Meeyoung Cha Sue Moon Chong-Dae Park Aman Shaikh Proc. of IEEE INFOCOM 2006 Speaker 游鎮鴻.
Courtesy Piggybacking: Supporting Differentiated Services in Multihop Mobile Ad Hoc Networks Wei LiuXiang Chen Yuguang Fang WING Dept. of ECE University.
Optimization-based Cross-Layer Design in Networked Control Systems Jia Bai, Emeka P. Eyisi Yuan Xue and Xenofon D. Koutsoukos.
Performance Comparison of Ad Hoc Network Routing Protocols Presented by Venkata Suresh Tamminiedi Computer Science Department Georgia State University.
COMP8330/7330/7336 Advanced Parallel and Distributed Computing Communication Costs in Parallel Machines Dr. Xiao Qin Auburn University
William Stallings Data and Computer Communications
Authors: Jiang Xie, Ian F. Akyildiz
Kevin Lee & Adam Piechowicz 10/10/2009
COS 561: Advanced Computer Networks
Data and Computer Communications
Towards Predictable Datacenter Networks
Presentation transcript:

QoS Routing Using Traffic Forecast - A Case Study of Time-Dependent Routing Yuekang Yang Chung-Horng Lung Dept. of Systems and Computer Engineering, Carleton University

SCE, Carleton University1 Scope Assumes that aggregated Internet traffic has a periodic predictable pattern Only the link capacity constraint and QoS constraints are considered in the discussion Link or node failure and traffic backup are not considered in this research Intra-domain QoS routing

SCE, Carleton University2 Outline Introduction Methodology Simulation Results of the State-Dependent Routing Mechanism Simulation Results of the Time-Dependent Routing Mechanism Conclusions

SCE, Carleton University3 QoS Routing QoS Routing is a part of Constraint-Based Routing CBR denotes a class of routing algorithms that base path selection on a set of constraints, in addition to the destination. If constraints imposed are QoS requirements, the associated routing is referred to as QoS routing. Other constraints in CBR could be administrative policies. Constraints in QoS Routing Bandwidth Delay Packet loss

SCE, Carleton University4 Challenges of QoS Routing Stability and Scalability When multiple resources are allocated and deallocated, high frequency of state updates is required to avoid instability and route flapping, but it does not scale well due to its high communication overhead for large networks. Robustness Routers always get state updates with delays, and there is no guarantee that resource information is accurate and up-to-date. Route computation and routing decisions should be robust enough based on imprecise states. Routing Cost Processing state updates, implementing techniques related to robustness issue, and conducting QoS routing (an NP-Complete problem) introduce considerable computational cost. Contrastively, QoS requests expect highly responsive services.

SCE, Carleton University5 Research Motivation Adapted from: “Long-Term Forecasting of Internet Backbone Traffic - Observations and Initial Models,” Proc. Conf. Comp. Commun., IEEE INFOCOM, March 2003.

SCE, Carleton University6 QoS Routing Using Traffic Forecast State-Dependent/Time-Dependent Mechanism A routing algorithm that is able to find the optimal route for every QoS request, without knowing either the history or the future traffic demand, is called state-dependent mechanism. A routing algorithm that has knowledge of the history or future traffic demand is called time-dependent mechanism. Three Questions Addressed in This Research Can traffic forecasting improve QoS routing performance? What if the forecast is not perfect? What kind of traffic forecast is required for QoS routing?

SCE, Carleton University7 Outline Introduction Methodology Simulation Results of the State-Dependent Routing Mechanism Simulation Results of the Time-Dependent Routing Mechanism Conclusions

SCE, Carleton University8 Simplification on QoS Constraints Delay and loss Latency formula from the hypothesis that each queue behaves as an M/M/1 queue of packets: D e is the processing and propagation delay on edge e, and c e is the link capacity of edge e. Both D e and c e could be considered as constant. x e is the total bandwidth required for all flows on edge e. It shows that new assigned QoS request can affect the delay of old flows. So the upper bounds of delay and loss for each edge are adopted to simplify the QoS routing problem in this work. It is justified by the fact that the measured loss probabilities and delay for the same QoS on different routers are of similar order. Hence, the number of hops along a route is considered as the constraint. Bandwidth Equivalent bandwidth is calculated according to burstiness, buffer size, flow peak/link ratio.

SCE, Carleton University9 Objectives in QoS Routing Two major objectives in QoS Routing Provide routing service with overall low network resource consumption. In other words, minimize the average number of hops that those routes traverse. Avoid overloading parts of the network while other parts are under- loaded. That is to minimize the maximum link utilizations. There are three reasons: Spare bandwidth is available at various parts of the network to accommodate unpredictable traffic requests. In case of link failures, smaller amounts of traffic will be disrupted and will need to be rerouted. Nodal router performance, such as delay, degrades dramatically when router’s load approximates its maximum capacity. In most cases, these objectives are not conformable with each other

SCE, Carleton University10 Performance Evaluation The Objective Function in Optimization c e is the link capacity on edge e. x e,k is the bandwidth of the kth flow on edge e. K represents the set of all flows in a network. The order n is a useful tool for ISPs to get more accurate results Based on this objective function, the optimal solution of routing can be calculated by the general gradient projection method. The value of this objective function can also be seen as an performance metric when different routing algorithms are compared.

SCE, Carleton University11 Outline Introduction Methodology Simulation Results of the State-Dependent Routing Mechanism Simulation Results of the Time-Dependent Routing Mechanism Conclusions

SCE, Carleton University12 Simulation Setup - Network #1

SCE, Carleton University13 Simulation Setup - Network #2

SCE, Carleton University14 Simulation Results - Network #1 SP: shortest path WSP: widest shortest path SWP: shortest widest path

SCE, Carleton University15 Simulation Results - Network #2

SCE, Carleton University16 Analysis of State-Dependent Mechanism Underlying Problems with State-Dependent Mechanism State-dependent mechanism faces all challenges mentioned before, no matter it is designed as Pre-Computation Routing or On- Demand Routing. In particular, pre-computing paths for all possible QoS requirements is extremely processor and memory consuming, as it has no idea about the future demands. Rearrangement dilemma To avoid rearrangement as much as possible, the performance of state-dependent routing algorithms degrades gradually as the traffic demand increases. It is difficult for state-dependent routing algorithms to reach performance optimal points when the traffic demand is at a moderate level.

SCE, Carleton University17 Rearrangement Dilemma Some flows have to be rearranged in terms of explicit routes and their assigned bandwidth. Rearrangement causes service disruption and significant signaling overhead to proceed with minimal disruption. The cost of rearrangement increases dramatically as rearrangement becomes frequent. Suppose a state-dependent mechanism can approximate an optimal point according to the current state. After a few minutes when new LSP requests come in, it has a high chance to rearrange existing user flows in order to keep the same level of performance. On the other hand, if fewer rearrangements are needed, it has to stay away from the optimal points. The dilemma partially stems from the ignorance of future traffic demand.

SCE, Carleton University18 Outline Introduction Methodology Simulation Results of the State-Dependent Routing Mechanism Simulation Results of the Time-Dependent Routing Mechanism Conclusions

SCE, Carleton University19 Design a Time-Dependent Routing Algorithm For a request of bandwidth bw from source node s to destination node d, with hop count constraint c, the pseudo code for the TDWSP is as follows: initialize expectedBw, plannedFreeBw, residueFreeBw if bw < expectedBw(s,d) { do WSP(s,d,c) on plannedTree if success { expectedBw(s,d) = expectedBw(s,d) - bw plannedFreeBw(s,d) = plannedFreeBw(s,d)- bw } else { do WSP(s,d,c) on residueNet if success { residueFreeBw(s,d) = residueFreeBw(s,d)- bw } } else { do WSP(s,d,c) on residueNet if success { residueFreeBw(s,d) = residueFreeBw(s,d)- bw }

SCE, Carleton University20 Traffic Demands

SCE, Carleton University21 Simulation Results Network #2, the first line of traffic demands WSP: widest shortest path TDWSP: time-dependent widest shortest path

SCE, Carleton University22 Simulation Results Network #2, the second line of traffic demands

SCE, Carleton University23 Simulation Results Network #2, the third line of traffic demands

SCE, Carleton University24 Advantages of the Time-Dependent Mechanism Now the challenges mentioned in the beginning of the presentation are reexamined in the context of time-dependent mechanism. For most LSP requests that the TDWSP expects, ingress router can make routing decision just based on preplanned local information. Only when unexpected requests come in, TDWSP uses WSP algorithm on the residue network. The existence of this fast execution path can mitigate all those challenges we discussed before. TDWSP has a lower standard deviation for the objective function value than WSP when the traffic load is below the level forecasted. In other words, TDWSP is less sensitive to the order of QoS requests than WSP. The modification from WSP to TDWSP could facilitate a distributed implementation.

SCE, Carleton University25 Analysis of the Time-Dependent Mechanism When the offered traffic load is not heavy in terms of the maximum traffic a network can handle, a time-dependent variation of the state-dependent routing algorithm could perform very close to the optimal boundary. When the offered traffic load is heavy, the performance of a time-dependent routing algorithm is degraded. If the demand forecast is at the moderate traffic level, a time- dependent routing algorithm could be insensitive to the position of the forecast. One fixed forecast may be effective enough to optimize the routing performance in a large area.

SCE, Carleton University26 Outline Introduction Methodology Simulation Results of the State-Dependent Routing Mechanism Simulation Results of the Time-Dependent Routing Mechanism Conclusions

SCE, Carleton University27 Conclusions The inherent disadvantages of state-dependent mechanisms, due to the lack of prediction of traffic. In order to avoid rearrangement to support QoS routing, the performance of SP, WSP and SWP is not optimized when the traffic load grows. The benefits of traffic forecasting in the time-dependent mechanism are further explained in light of its positive impact on routing performance. TDWSP outperforms WSP under almost all traffic conditions in both networks, #1 and #2. TDWSP greatly relaxes the burden of QoS database synchronization among routers, and hence eases the challenges of scalability, robustness and routing cost.

Backup slides SCE, Carleton University28

SCE, Carleton University29 The Curvature of the Objective Function The objective function is: This objective function is a sum of link costs. The diagram shows the link cost increases with the link utilization.

SCE, Carleton University30 Effect of the Curvature of the Objective Function

SCE, Carleton University31 Effect of the Curvature of the Objective Function

SCE, Carleton University32 Simulation Results Network #1, the first line of traffic demands

SCE, Carleton University33 Simulation Results Network #1, the second line of traffic demands

SCE, Carleton University34 Simulation Results Network #1, the third line of traffic demands

SCE, Carleton University35 The Solvable Traffic Demand Space For instance, a four-node network has the following traffic matrix: The traffic matrix can also be represented in vector form when the topology information is not relevant to the discussion. Imagine a space that consists of the end-points of all legitimate vector d, and call it the traffic demand space. From the routing point of view, in this research we define a solvable traffic demand space as a subset of the traffic demand space with at least one feasible routing solution, no matter whether the feasible routing solution is optimized or not.

SCE, Carleton University36 Implications The unsymmetrical insensitivity of the traffic demand does suggest that the extreme value of the daily traffic orbit might be the focus of the traffic forecast. If the routing performance under peak traffic is good, then the routing performance during the rest of the day should be fine. The size and shape of the solvable traffic demand space is determined if the topology of a network has been finalized. However, the solvable traffic demand space is not very large, given a topology. The mapping from the solvable space to specific routing solutions is the job of the QoS routing algorithms. If our focus is turned toward the solvable space that is not large in size, then it seems reasonable to introduce the time-dependent routing mechanism.